The phrase ‘extraordinary times call for extraordinary measures' may feel like homespun wisdom, but it became risk management policy at Societe Generale Corporate & Investment Banking (SG CIB) last year, helping the bank ride out two major market events.
Bespoke stress tests gave SG CIB a clearer idea of its exposures heading into both Brexit and the autocallables drama of January and February. It then used the results to cut its risk well in advance of the events – more than halving foreign exchange gamma ahead of the UK's referendum, for example. That helped shield the bank from substantial trading losses and hedging costs.
In the case of the Brexit vote, the process started after top brass at the bank made their risk management expectations pointedly clear – something they have been doing more regularly in recent years.
"We receive these kinds of instructions on a regular basis now. But the instructions were more specific this time," says Bruno Gaussorgues, global head of market risk at SG CIB. "The board and the senior management were not keen on losing a large amount of money on this day, or making a large amount of money. They were focused on being insensitive to event risk."
In theory, that would enable the bank to perform more effectively as a market-maker, and capture more client flow in the event of a surge in trading. In practice, it was exactly the right strategy.
In the days before the referendum, dealers had an opportunity to buy relatively cheap out-of-the-money currency options, but misplaced confidence in a Remain victory encouraged some to leave a good portion of downside risk open. These banks – and many clients – then had to cover their positions as polling results started filtering through.
This time we went further and looked at so many different scenarios that we all ended up having a common view of what was the residual risk appetite in order to satisfy the last-minute client orders
Gonzague Bataille, SG CIB
It was a bad time to be in that position. Liquidity dried up during early trading, and spreads widened by multiples for a range of asset classes. Although not life-threatening, the dislocations following the vote hurt many market participants.
But not SG CIB. "One or two months before the vote, we started running specific stress tests on UK rates, sterling and UK equities. Usually our market stress tests are multifactor stress tests, but we wanted to look at a lower number of factors to understand the position well," says Gaussorgues.
That meant the team had to carry out the time-consuming task of analysing the contributions made by those individual risk factors in a new multi-factor stress test built specifically for Brexit, where simultaneous shocks were applied to a group of risk factors, reflecting particular stressed scenarios.
On the back of the results – and in light of the board's instructions on risk-taking – SG CIB decided to bring down its exposure.
"We reduced negative gamma on both forex and rates. At the end, we entered the day of the referendum result with very residual positions both on market and counterparty credit risk, with the ability to trade and answer to client requests," says Gaussorgues.
That may sound simple enough. In fact, the tests were unusual in a number of ways. Because the bank was trying to assess its exposure to movements at a particular future date, it needed to take into account the likely make-up of its portfolio at that point. That meant considering the impact of options expiries.
"Brexit happened just after the quarterly expiry on the Euro Stoxx options. It is usually a very significant expiry for a lot of products that are traded either as a hedge or directly with clients," says Francois Levy-Bruhl, head of trading risk at SG CIB. "The difficulty was not only about withdrawing all these options that were maturing, but also unwinding any hedge you have on the day the options mature. Typically, you can have three-month futures against something that will expire in one week. It is clearly something that could have inverted the figures."
The challenge was a new one for the bank, says Gonzague Bataille, SG CIB's deputy head of market risk – the test it designed had to look at a host of scenarios that might play out, starting from the day before the vote, while also factoring in the expiring options.
"There was a precise date at which you had to run the simulation. You have to filter out all the different transactions that are going to expire and their associated deltas," says Bataille. "At least on our side, it was the first time we looked into this. We have the classical skew or gamma profile on different expiries, but this time we went further and looked at so many different scenarios that we all ended up having a common view of what was the residual risk appetite in order to satisfy the last-minute client orders."
Another key component of the test was cross risk or wrong-way risk, which is the risk of an increase in counterparty risk due to adverse movement in exposures. SG has been monitoring this risk since 2012 on its biggest – or shakiest – counterparties, and taking measures to hedge or reduce positions when certain metrics related to systemic risk and concentration risk are breached. This analysis was also carried out for Brexit.
"We did our own view of our positions facing our most sensitive clients and some UK clients. With Brexit, it was to check that a massive sterling move or massive equity move would not dramatically change our counterparty credit risk with these clients," says Levy-Bruhl. For the Brexit scenario, the analysis flagged up six counterparties, but SG CIB decided each was strong enough that no action was required.
Follow-ups done over the weekend after the vote showed the Greeks from the stress test were roughly in line with trading estimates based on activity on Friday.
"We had dedicated teams during the weekend following Brexit day to run the most important metrics in advance. We wanted to be sure that on Sunday evening we would be able to know where we stand in terms of market risk and be able to communicate that to our management and regulators," says Levy-Bruhl.
Brexit was not the only storm the bank weathered using stress tests.
In early 2016, the autocallables market took a beating as global indexes nosedived, the most notable being that of the Hang Seng China Enterprises Index, which is the underlying exposure for a large volume of outstanding Korean equity-linked autocallables. By February 12, the index was down 22% for the year, and had dipped below the 8,000-level where downside barriers were said to be clustered.
Thanks to a newly built monitoring system, SG was able to manage this episode better than some of its peers. Risk managers and traders at the bank jointly developed the volatility convexity adverse stress test (VCAST), an early warning system to avoid passive breaches in autocallables, which the bank rolled out in June 2015.
Autocallables typically have an upside barrier and a downside barrier, with the latter consisting of a knock-in put option, sold by the investor to the issuer. If the upside barrier is breached, the whole structure knocks out and investors are repaid their principal plus a coupon. If the index moves towards the downside barrier, the put held by the dealers becomes more valuable – they are increasingly long vega, and need to sell puts to remain flat. The exposure changes again if the downside barrier is breached and the options are triggered. At this point, the investors' principal is at risk, but dealers also have to buy back volatility using variance swaps and options to flatten their exposures.
The process of hedging and re-hedging during a sliding market can be very expensive, given that all autocallable issuers have the same exposures. In one such episode in September 2015, dealers collectively lost an estimated $300 million.
"The cost of hedging increases rapidly as you try to cover your positions that are increasing at the same time. It is a significant position for us, and we think there is limited awareness in the market about the liquidity issue," says Bataille.
Using VCAST, which churns out metrics that signal possible breaches, SG CIB was able to ride out last year's stress without pain.
VCAST consists of a carry stress test, in addition to a standard spot stress test. Losses related to the latter may be recovered if the stress is short-lived, whereas the former looks at carry P&L – or moves that cannot be recovered – and monitors long-term second-order risks across all of the bank's autocallable books. The carry stress test is built to look at potential P&L losses in scenarios where, over an extended period of time, the spot-volatility correlation and the volatility of volatility end up being different to those in model calibrations, and the volatility smile moves adversely. The first two effects are captured by the vanna and vomma sensitivities, respectively.
When the spot goes down, there is a potential for not only mark-to-market losses, but significant hedging losses as dealers rush to put on hedges in the same direction in a market that is increasingly becoming illiquid. Vanna and vomma help indicate the stability of the hedging portfolio during such market stresses by capturing the non-linear movement of the products during a liquidity crunch.
This framework forces us, in a way, to anticipate the deterioration of market conditions and put a number of protections in place rather than wait for a passive breach – in which case we won't be able to get this protection in the market
Arie Boleslawski, SG CIB
"It is not really a mark-to-market stress where you bump market parameters and revalue your positions. What we want to assess is how far the position can passively move in a really adverse scenario, where the market as a whole will not find proper liquidity. In terms of liquidity horizon, it is much longer than the time horizons we use for the rest of our positions," says Bataille.
The monitoring of risks using VCAST allows SG to not only quantify these second-order risks, but also to show how robust the bank's volatility hedges are over time in very adverse spot-volatility scenarios. Three different VCASTs are computed in order to set risk limits – one at the current market condition, a second at equity spot minus 15%, and the third at equity spot minus 25%. The breach of these limits informs what hedging strategy should be followed.
"We know if the market starts to go down, liquidity will disappear and therefore the position will change passively without being able to re-hedge. We use what we call the shifted VCAST to ensure we are well covered, even if the market is 25% down," says Levy-Bruhl.
For SG, most of the macro hedges were completed before the start of the autocallable turbulence and the bank remained within its risk appetite – as measured by the VCAST metric – throughout the episode. Traders were given three months to execute the hedging programme.
"This framework forces us, in a way, to anticipate the deterioration of market conditions and put a number of protections in place rather than wait for a passive breach – in which case we won't be able to get this protection in the market," says Arie Boleslawski, the Paris-based deputy head of trading at SG CIB.
The week on Risk.net, August 4–10Receive this by email